A Study of the Cooperation Control of Two Adjacent Intersections Based on N.B.S. Game Q-Learning Algorithm

A new control algorithm - N.B.S.Game Q-Learning was introduced to solve the cooperation control of two adjacent intersections in this paper. N.B.S.Game was denoted as two-player cooperation game with Nash bargaining solution. Based on game Q-learning algorithm, that the game theory was combined with the Q-learning realized by BP neural network and the game solution was regarded as the basis of taking the strategy selecting of Q-learning, the N.B.S.Game Q-learning algorithm was just put forward. Because the traffic signal cooperation control problem for two adjacent intersections belonged to the two-player general sum cooperation game form, the Nash bargaining solution method was applied to obtain the optimal portfolio strategy to ensure the maximization of the overall benefit. The simulation result by Paramics has showed the control performance of the N.B.S.Game Q-learning algorithm is far better than fixed time control in heavy traffic flow condition and the control strategy can adapt to the variable traffic environment.